You've got an 80-page course packet, two PDF reports, a pile of notes you scribbled in a hurry, and a two-hour YouTube video. You know the answer you need is somewhere in there, but you don't have three evenings to spend hunting for it. NotebookLM is Google's tool built for exactly this moment: you drop in your documents, you ask your questions in plain English, and the tool answers while citing the exact passage each piece of information came from.
The difference with ChatGPT fits in one sentence: NotebookLM only answers from what you've given it. It won't invent a number it picked up somewhere else, and when the information isn't in your sources, it tells you so.
In the next few minutes, you'll see how to create your first notebook in five steps, which features are actually worth your time (including the audio summary), what limits apply to the free version, and when a different tool would serve you better.
What NotebookLM actually is
NotebookLM is a free research assistant from Google that works only from the documents you give it, with clickable citations back to the original passage.
The idea is simple. You create a notebook, you add up to 50 sources to it, and then you chat with the whole set the way you'd chat with a colleague who had read everything. Every answer shows little numbers: you click one, and the exact passage opens in the source document. That traceability is what changes the way you work, because you can verify something in two seconds instead of just trusting it.
The supported formats cover almost everything a beginner is likely to have on hand:
- PDFs (including entire books), .txt files, and Markdown
- Google Docs and Google Slides from your Drive
- web pages (you just paste the address)
- YouTube videos, via their automatic transcripts
- audio files (a meeting recording, a recorded lecture)
- copy-pasted text, for your loose notes
Each source can hold up to 500,000 words, which is roughly five thick textbooks. You need a Google account, and that's it: NotebookLM runs straight in your browser, with nothing to install.
The technical term behind all this is grounding: the model is forced to draw its answers from a closed corpus. The practical consequence for you is far fewer made-up errors than with a general-purpose assistant, but also zero outside knowledge. Ask it for tomorrow's weather and you'll get nothing.
Three ways to learn from your own documents
The right approach depends on your goal: building AI skills in a structured way, digesting a specific set of documents, or asking a quick question about one file.
Method 1: a guided path with the Skilzy AI Lab
NotebookLM answers the question “what's in my documents” very well. What it doesn't do is teach you to write good prompts, produce visuals, edit a video, or automate a repetitive task. If your real goal is to know how to use AI day to day, you'll save time with a path that walks you through it step by step.
That's the idea behind Skilzy: more than 15 programs covering image creation, video (UGC, ads, faceless channels, avatars, voiceovers), music, automation with n8n, prompt writing, and AI-assisted development. The Skilzy AI training programs are built around a built-in AI Lab: you use the real tools right inside the platform, credits included, without stacking up five separate €20-a-month subscriptions. Two state-recognized, fundable certifications round it out (RS7439 for AI content marketing, RS6792 for AI and sales).
Method 2: NotebookLM
This is the tool to pick when you have a closed, clearly identified corpus: a training curriculum, an employee handbook, ten scientific papers, the meeting notes for a project. Free, precise, with citations. Its limits: daily quotas and total ignorance of anything that isn't in your sources.
Method 3: ChatGPT, Claude, or Gemini with attachments
For a one-off question about one or two files, these assistants get it done faster, with no notebook to set up. They're also better at drafting a long piece of text from the document. Beyond a handful of files, their reliability drops and citations become less consistent. If you're torn between them, our comparison of the free versions of ChatGPT, Claude, and Gemini breaks down the differences.
| Method | Best for | Price | Main limitation |
|---|---|---|---|
| 1. Skilzy (programs + AI Lab) | learning to use AI end to end, with certification | Monthly subscription | it's not a search engine for your PDFs |
| 2. NotebookLM | digesting a specific set of documents | free with a Google account | daily quotas, nothing outside your sources |
| 3. ChatGPT, Claude, Gemini | a quick question about 1 to 3 files | free with limits | less reliable citations at high volume |
Create your first notebook in 5 steps
Budget about ten minutes to end up with a working notebook and your first sourced answer.
1. Open NotebookLM and sign in. Head to notebooklm.google.com with your Google account, then click “Create notebook.” Give it a clear name, something like “Employment law revision”: you'll have several of them within a month.
2. Add your sources. Drag your PDFs into the window, paste in web page or YouTube addresses, pick documents from your Drive. One tip that prevents 80% of disappointments: one notebook equals one topic. Mixing your thesis, your tax return, and your dinner recipes produces vague answers.
3. Set the output language. In the notebook settings, choose the language you want answers in. Without that setting, the tool sometimes replies in a different language from the one you wrote in, especially for audio summaries.
4. Ask real questions. Vague questions (“summarize this”) get vague answers. Try these instead:
- “Summarize this course in 10 numbered points, with a citation for each one.”
- “What are the three contradictions between the 2024 report and the 2025 report?”
- “Write me 15 practice questions with an answer key, from easiest to hardest.”
- “Explain the concept of a sliding scale as if I were 12 years old, using only my sources.”
Make a habit of clicking one or two citations to check them. That reflex is what separates serious use from naive use.
5. Keep what's good. Pin the useful answers in the notebook's notes. You can then turn a note into a source, which lets you build your own synthesis and question it again later.
The features that genuinely save you time
Beyond the chat, four features turn your documents into ready-to-use learning material.
The audio summary is the best known. NotebookLM generates a podcast-style conversation between two synthetic voices from your sources. Expect about ten minutes of listening for a large document. English is available, along with more than 70 other languages, and you can even interrupt the discussion to ask a question out loud. Perfect for reviewing while walking or driving.
The mind map lays out the tree of concepts in your notebook. You click a branch, and the question is automatically sent to the chat. Very effective for seeing the structure of a subject you're new to.
The reports produce, in one click, a study guide, a briefing doc, a timeline, or an FAQ based on your sources. The study guide is especially handy: definitions, open questions, a short quiz.
The video summary generates a narrated presentation, useful when you need to explain the content to someone else rather than learn it yourself.
Two side features are worth a try: source discovery, which suggests relevant web pages to add to your notebook, and sharing, which lets a study partner work on the same notebook. To sort through these options and the other AI tools out there right now, our guide to learning AI as a complete beginner gives a clear order of priority.
Four concrete use cases
NotebookLM becomes genuinely useful the moment you have more documents than time.
Student in exam season. You drop in the professor's slides, your notes, two textbook chapters, and the video of the online lecture. You generate a study guide, then ask for 20 practice questions. You move from passive reading to active practice, which makes things stick better.
Employee starting a new role. Internal procedures, meeting notes, technical documentation: one notebook per project, and you can finally ask “who decided what, and when?” without bothering three colleagues.
Parent dealing with homework. You add the textbook chapter and ask for explanations pitched at your child's age, or extra exercises. Use with care: the goal is to help them understand, not to hand over the answers. Our article on how to help your child with homework using AI covers the guardrails.
Freelancer or small-business owner. A 60-page spec, a request for proposal, a set of terms and conditions: you ask for the obligations, the deadlines, and the penalties, with citations. The time saved is measured in hours.
Limits, quotas, and privacy
NotebookLM is free but capped, and a few precautions are in order before you upload sensitive documents.
As of this article's publication date (October 2026), the free version allows roughly 100 notebooks, 50 sources per notebook, about fifty chat questions per day, and a handful of audio summaries per day. Google's paid subscription multiplies these ceilings by about five and raises the limit to 300 sources per notebook. These numbers change regularly, so check the current values in the official NotebookLM help center.
Three technical limits worth knowing:
- a scanned PDF without text recognition works poorly, because it's made of images rather than text; convert it first
- the tool isn't built for drafting a long original document, where a general-purpose assistant does better
- YouTube transcripts contain errors, especially with proper nouns and numbers
On privacy, Google states that files uploaded to the consumer version are not used to train its models. It's still a service hosted on a third party's terms: avoid uploading clients' personal data, medical records, or documents under an NDA without clearing it first. The rule is simple: if you wouldn't email it to an outside vendor, don't put it in a notebook.
One last point people often forget: NotebookLM mirrors the quality of your sources. A shaky corpus produces shaky answers, delivered with total confidence. Verifying through citations isn't optional, it's the instruction manual.
Key takeaways
NotebookLM is the most direct way to turn a pile of documents into something you can actually question, with verifiable citations and an audio summary you can listen to. It costs you zero euros and ten minutes of learning. Start with a single notebook, on a topic that's genuinely occupying you this week, and ask it five specific questions. If you then want to go further than reading your own files, meaning creating, automating, and producing with AI, a structured path plus access to the real tools will move you forward far faster than piling up scattered tutorials.
